hi,
Thanks for the fixes in version 0.8.6;) However I'm very curious about
changing gen_event:notify to sync_notify in log_manager.erl. Actually
it works quite good - no memory crashes and it works faster - during 5
minute test I created 38000 logging processes compared to 26000 using
older version of log4erl. Could somebody explain to me how it is
possible that introducing sync logging caused Erlang to run faster?
regards,
Kuba
On 8 Kwi, 09:30, Ahmed Ali <
ahmed.naw...@gmail.com> wrote:
> Hi,
>
> Please check thishttp://
log4erl.blogspot.com/2009/04/log4erl-issue-when-high-load.html.
>
> Regards,
>
> Ahmed
>
> 2009/4/7 Ahmed Ali <
ahmed.naw...@gmail.com>:
>
> > Hi Kuba,
>
> > I forgot to mention that the best thing to do right now is to raise
> > logging level to something like warn and shorten log messages.
> > However, this is not a solution.
>
> > Regards,
>
> > Ahmed
>
> > 2009/4/7 Ahmed Ali <
ahmed.naw...@gmail.com>:
> >> Hi Kuba,
>
> >> Obviously, the crash is because many requests comming to file appender
> >> and requests accumulate waiting for IO to finish until process inbox
> >> takes a lot of memory and then erlang crash. A case study about this
> >> is in [1].
>
> >> Unfortunately, this will almost always happens in a concurrent
> >> language like erlang and log4erl, and error_logger for that matter,
> >> will always suffer from this.
>
> >> I've actually been thinking about this for sometime and I think the
> >> best option is to add persistent queue with something like memcacheq
> >> as the backend. This will slow down logging but it will not cause an
> >> "out of memory" crash because of many requests. If anyone has a better
> >> idea, please let me know.
>
> >> [1]
http://blogtrader.net/dcaoyuan/entry/a_case_study_of_scalable
>
> >> Regards,
>
> >> Ahmed
>
> >> 2009/4/7 odi <
kuba.od...@gmail.com>:
>
> >>> Kliknijhttp://
groups.google.com/group/log4erl/web/file-appender-crash-during...